Classification of Physical Activities Based on Photoplethysmography Signals from a Wearable Device

Abstract

Recent advances in wearable technology have enabled the users to monitor their physical and physiological states from sensors in real time. In this study, we have proposed a computational method to detect different types of physical exercises only from photoplethysmography (PPG) signals obtained by a wrist-type wearable device. Our method was composed of feature extraction from PPG and classification using a linear discriminany analysis algorithm. Using the developed method, we could classify two different types of exercises in an individual with accuracy of 78% on average. Our proposed method may be useful to monitor the physical activities of the user and to provide customized u-healthcare services for individuals.